On Design of a Three- Class ECG Classifier

Objectives: Manual analysis of ECG, specifically the ambulatory type, is tedious and time consuming, hence automation is desired. In this paper the authors have focused on the design of a three class ECG classifier using feature extraction and artificial neural networks. Once feature extraction is done, ANNs can be trained to classify the patterns reasonably accurately. Arrhythmia is one such type of abnormality detect able by an ECG signal. The three classes of ECG signals are Normal, Fusion and Premature Ventricular Contraction (PVC). The task of an ANN based system is to correctly identify the three classes, most importantly the PVC type, this being a fatal cardiac condition. Methods: The ECG data is taken from MIT-BIH Arrhythmia database. Fifty-five different feature extraction schemes are examined, along with a compact set of statistical morphological features and a reasonably accurate and fast classifier is designed. These feature extraction techniques coupled with ten morphological features have not been collectively studied and compared in literature so far. Findings: Multilayer perceptron with momentum learning rule is found to be the best classifier topology, while the best performing feature extraction schemes are: bior2.2, coif1, db9, rbior1.1, sym2, DCT and PCA. Application/Improvements: The reported findings can be effectively used for automated ECG arrhythmia classification for rapid analysis by a cardiac specialist thus saving time and arriving at a quick and reliable diagnosis. A similar approach can be used for designing an ECG classifier for more number of arrhythmia conditions.